Post

@sundogplanets tbf not wanting to sound mean, but generally there is a severe lack of funding and appriciation for science computing and the assumption of oh it will exist and people will maintain it with free money is just a lie. But getting people convinced they need to higher and properly compensate staff for it is practically impossible. (And astronomy is definitely one of the worse ones in this)

@sundogplanets
Even when grants exist, as of late last fall purchasing is very hard and you get less than half of what the money should cover. All due to AI datacenters with way too much money outbidding us at every step of the chain, from DRAM dies to full servers.

The official strategy for LHC computing now is to use lots of duct tape and hope the current hardware lasts until the bubble bursts.

@sundogplanets yes, at some point there's an element of faith.

The question is: how long will we be able to afford simply maintaining things as they are...?

@sundogplanets I didn't even know that's a thing. What's in it? Who runs it? Who gets access? I have so many questions...

@sundogplanets
I used to work for a company trying to address that; a grid computer that runs on scavenged cycles from enrolled servers and desktops. Part of their intro deck was that Canadian super compute sucks intersectionally so it might be way better to use your own resources.

If your workload is embarrassingly parallel and is python or JavaScript, their platform might be of interest?

@sundogplanets Yep. And when was the last time the military had to have a bake sale to buy the latest people killing tech?

@sundogplanets even big systems like $dayjob feel like we're relying on hope and duct tape when it comes to persuading funding bodies that yes, science investment IS important and requires HPC.

@sundogplanets "The astronomy cloud computing system is down."

Forgive me for the knee-jerk reaction about how astronomy and clouds don't mix....